NCT07406269

Clinical Evaluation of AI Decision Support for Early Rehabilitation After Surgery

Evaluation of Clinical Effectiveness and Implementation of an Artificial Intelligence Based Decision Support Tool That Guides Early Rehabilitation After Gastrointestinal and Oncology Surgery

• Status: Not Yet Recruiting
• Phase: N/A
• Study type: Interventional
• Enrollment: 103 (estimated)
• Sponsor: Erasmus Medical Center · Academic / Other
• Sex: All
• Age: 18 Years
• Healthy volunteers: Not accepted

Summary

After gastrointestinal or oncology surgery, it can be difficult to determine when a patient is ready to safely begin early rehabilitation or move toward discharge. Delays may prolong hospital stay, while premature decisions may increase risks. This study evaluates an artificial intelligence (AI)-based decision support tool that analyzes routinely collected hospital data to identify patients who are likely ready for early rehabilitation and discharge planning after surgery. The tool provides a simple yes/no output to support clinicians in their decision-making. The AI tool does not replace clinical judgment. Treating physicians remain fully responsible for all care decisions. The purpose of this study is to examine how well this tool performs in clinical practice and how it can be safely and effectively implemented to support postoperative care.

Detailed description

Patients who undergo gastrointestinal or oncology surgery often require careful monitoring after their operation. During the days following surgery, healthcare professionals assess many factors, such as vital signs, laboratory results, recovery progress, and the need for hospital-based treatments. Based on this information, decisions are made about when patients can safely start early rehabilitation or move toward discharge planning. In this study, researchers are evaluating an artificial intelligence (AI)-based decision support tool designed to assist clinicians with these decisions. The tool analyzes routinely collected information from the electronic patient record, including demographic data, type of surgery, vital signs, laboratory values, and medication information. Using these data, the system provides a simple yes/no output indicating whether a patient is likely ready for early rehabilitation and discharge planning on the second day after surgery. The AI tool is advisory only. It does not make treatment decisions and cannot initiate any actions. The treating physician always reviews the patient's condition independently and makes the final decision about care, rehabilitation, and discharge planning. The study focuses on two main aspects: 1. How accurately the AI tool identifies patients who are ready for early rehabilitation and discharge planning. 2. How the tool can be safely and practically integrated into everyday clinical workflows. Participation in this study does not change the standard of care. All patients continue to receive routine postoperative care according to existing hospital protocols. The AI tool serves solely as an additional source of information for clinicians. Patient data used by the AI system are processed within secure hospital systems and handled in accordance with data protection regulations. No additional tests or procedures are required specifically for this study. The results of this study may help improve postoperative care by supporting timely rehabilitation and discharge planning, potentially reducing unnecessary hospital stays while maintaining patient safety.

Conditions

• Oncologic Surgery
• Gastrointestinal Cancers

Interventions

Type	Name	Description
DEVICE	DESIRE: AI-Based Clinical Decision Support for Postoperative Rehabilitation Planning	The intervention consists of the clinical use of a locked, non-adaptive artificial intelligence (AI)-based clinical decision support system (DESIRE) that analyzes routinely collected electronic health record data to predict, on postoperative day 2, the risk that a patient will require hospital-specific interventions after gastrointestinal or oncological surgery. The system automatically extracts demographic, perioperative, vital sign, laboratory, and medication-related variables and generates a binary (yes/no) output indicating whether the patient is likely to be at low risk for requiring additional hospital care. A predefined conservative threshold is used to identify patients eligible for early rehabilitation.

Timeline

• Start date: 2026-06-01
• Primary completion: 2027-06-01
• Completion: 2027-07-01
• First posted: 2026-02-12
• Last updated: 2026-02-12

Locations

1 site across 1 country: Netherlands